Berlin 2024 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

KFM: Fachverband Kristalline Festkörper und deren Mikrostruktur

KFM 9: KFM Poster Session

KFM 9.22: Poster

Monday, March 18, 2024, 18:00–20:00, Poster E

Using convolutional networks to predict the long term evolution of a multiphasic material — •Shing Wan and Nigel Clarke — Department of Physics and Astronomy, University of Sheffield, Sheffield, UK

Understanding the solidification and morphology of alloys has gathered resurgent interest with the recent advancements in metallic based additive manufacturing and low dimensional materials such as graphenes. The physical property of a material such as flexibility, tensile strength among others depends on the morphology of the material. The morphology can be inferred using the distribution of interfaces, regions between grains/phases, throughout the material.

Morphological properties of a material acquired via experiments can be represented as an image. This transforms the task of morphological evolution prediction to a frame prediction /generation task similar to those used in video games.

We are developing a machine learning approach to predicting microstructure evolution. Our methodology is based on a convolutional autoencoder in combination with a convolutional Long Short Term Memory. To investigate the effectiveness of the machine learning model, we used grain growth evolution of a multiphase alloy system, simulated using multiphase field theory.

Keywords: Phase field model; Machine Learning; Microstructure Evolution